1515792 - New URL Tab crashes after resume from Hibernate

Status: RESOLVED DUPLICATE of bug 1520760

(Core :: Graphics: ImageLib, defect, P3)

Description

[Terrox]

User Agent: Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:64.0) Gecko/20100101 Firefox/64.0

Steps to reproduce:

Firefox 64 Bit, version 64.0. Nvidia drivers 417.35 last 2 versions same result but I don't think Nvidia driver is the cause.

Firefox Settings
Use recommended performance settings - OFF
Use hardware acceleration when available - ON
Current Process Limit - Default (4)

Have some website tabs open.
Hibernate.
Resume from Hibernate
New tab, enter a URL

Work Around is set Current Process Limit - 1
No new tab crash if Current Process Limit = 1



Actual results:

Gah new tab crashed. 100% of the time except for some existing tabs and URLs like google.com do not crash.
No Crash Report generated as Firefox itself didn't crash?

Maybe related to this on OSX https://bugzilla.mozilla.org/show_bug.cgi?id=652828


Expected results:

New URLs in new Tabs should not crash after resume from Hibernate when using Hardware Processes more than 1

Comment 1

[Marcia Knous [:marcia]]

Terrox: If you have a crash report in about:crashes, can you please paste it here? Thanks!

Comment 2

[Terrox]

(In reply to Marcia Knous [:marcia - needinfo? me] from comment #1)
> Terrox: If you have a crash report in about:crashes, can you please paste it
> here? Thanks!

I do not, these tab crashes do not create crash reports as far as I know.

Comment 3

[Terrox]

I should specify that it is new uncached URLs which crash the tab, already visited URLs will not crash a new tab after resume from hibernate for me.

Comment 4

[Terrox]

Just had another crash with 1 cpu thread, so my work-around doesn't work. It seemed to only crash when I scrolled to a part of the page with uncached images and it can't load the images. So maybe my cache settings are weird?

Comment 5

[Terrox]

more info on my fidings. It doesn't crash when uncached GIF or PNG is loaded but does crash when uncached JPG is loaded. testing this on Reddit and my own websites where I know they content types being loaded.

Comment 6

[Terrox]

I believe I have narrowed it down to a BIOS setting.

Suspect: ASUS BIOS seting "limit CPUID max" = enabled = tab crash loading JPEG data, no firefox crash report is generated.
Intel i5 CPU 2500K, 8GIG RAM, Windows 7 64bit, in both 64 and 32bit Firefox.
Tested different Windows 7 OS install too, same crash situation.

It also crashes after Sleep resume, I mainly use Hibernate which causes the tab crash. My system uses Hybrid sleep.

I can consistently get this crash to happen when I change the BIOS setting "limit CPUID max" to Enabled. I've tested a lot of other stuff and am fairly certain it isn't any of the other stuff. I will leave the BIOS setting Disabled so it shouldn't crash and see if I get another crash over the next few days.

I think "limit CPUID max" = Disabled is the default BIOS setting so it would be fairly uncommon. I only changed the settnig because I tinker with things and it says "Disabled for XP" and that sounds like it is safe to Enable for whatever reason. :|

Comment 7

[Hani Yacoub, Desktop QA]

I am going to assign the "Core: Graphics" component for it.

Comment 8

[Terrox]

No crashes since changing BIOS settings for 19 days.

Comment 9

[dwoldrich]

I also am seeing this bug since 64 landed, exactly as described with tabs immediately crashing during load. I have a very similar Windows 7 configuration as @Terrox except I have a Radeon display. No crash reports are logged.

Will see if I also have Limit CPUID max = Enabled and report back if I see similar success coming out of hibernate over the coming days.

Comment 10

[dwoldrich]

I also had Limit CPUID max = Enabled, I switched it to Disabled, and that was the only configuration change I will make. (Mine also says "Disabled for XP" in the description, so @Terrox and I may be long lost identical bios twins.)

Comment 11

[Andrew Osmond [:aosmond] (he/him)]

This sounds like bug 1520760 to me.

Comment 12

[Andrew Osmond [:aosmond] (he/him)]

I'll block it on the bug, so that we can review when the fix lands.

Comment 13

[Andrew Osmond [:aosmond] (he/him)]

Would you be able to try the latest nightly build and confirm whether or not the problem is fixed? Thanks!

Comment 14

[Terrox]

(In reply to Andrew Osmond [:aosmond] from comment #13)

Would you be able to try the latest nightly build and confirm whether or not the problem is fixed? Thanks!

Not fixed. @dwoldrich can you also test?

Comment 15

[Terrox]

I tested in 66.0a1

Comment 16

[Andrew Osmond [:aosmond] (he/him)]

We should have fixed this as part of bug 1520760.